1 /* 2 * Copyright © 2016 Intel Corporation 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice (including the next 12 * paragraph) shall be included in all copies or substantial portions of the 13 * Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 21 * IN THE SOFTWARE. 22 * 23 */ 24 25 #ifndef SHADER_INFO_H 26 #define SHADER_INFO_H 27 28 #include "util/bitset.h" 29 #include "shader_enums.h" 30 #include <stdint.h> 31 32 #ifdef __cplusplus 33 extern "C" { 34 #endif 35 36 #define MAX_INLINABLE_UNIFORMS 4 37 38 struct spirv_supported_capabilities { 39 bool address; 40 bool atomic_storage; 41 bool demote_to_helper_invocation; 42 bool derivative_group; 43 bool descriptor_array_dynamic_indexing; 44 bool descriptor_array_non_uniform_indexing; 45 bool descriptor_indexing; 46 bool device_group; 47 bool draw_parameters; 48 bool float32_atomic_add; 49 bool float64; 50 bool float64_atomic_add; 51 bool fragment_shader_sample_interlock; 52 bool fragment_shader_pixel_interlock; 53 bool generic_pointers; 54 bool geometry_streams; 55 bool image_ms_array; 56 bool image_read_without_format; 57 bool image_write_without_format; 58 bool image_atomic_int64; 59 bool int8; 60 bool int16; 61 bool int64; 62 bool int64_atomics; 63 bool integer_functions2; 64 bool kernel; 65 bool kernel_image; 66 bool literal_sampler; 67 bool min_lod; 68 bool multiview; 69 bool physical_storage_buffer_address; 70 bool post_depth_coverage; 71 bool ray_tracing; 72 bool ray_query; 73 bool ray_traversal_primitive_culling; 74 bool runtime_descriptor_array; 75 bool float_controls; 76 bool shader_clock; 77 bool shader_viewport_index_layer; 78 bool stencil_export; 79 bool storage_8bit; 80 bool storage_16bit; 81 bool storage_image_ms; 82 bool subgroup_arithmetic; 83 bool subgroup_ballot; 84 bool subgroup_basic; 85 bool subgroup_quad; 86 bool subgroup_shuffle; 87 bool subgroup_vote; 88 bool tessellation; 89 bool transform_feedback; 90 bool variable_pointers; 91 bool vk_memory_model; 92 bool vk_memory_model_device_scope; 93 bool float16; 94 bool amd_fragment_mask; 95 bool amd_gcn_shader; 96 bool amd_shader_ballot; 97 bool amd_trinary_minmax; 98 bool amd_image_read_write_lod; 99 bool amd_shader_explicit_vertex_parameter; 100 bool amd_image_gather_bias_lod; 101 102 bool intel_subgroup_shuffle; 103 bool intel_subgroup_buffer_block_io; 104 }; 105 106 typedef struct shader_info { 107 const char *name; 108 109 /* Descriptive name provided by the client; may be NULL */ 110 const char *label; 111 112 /* Shader is internal, and should be ignored by things like NIR_PRINT */ 113 bool internal; 114 115 /** The shader stage, such as MESA_SHADER_VERTEX. */ 116 gl_shader_stage stage:8; 117 118 /** The shader stage in a non SSO linked program that follows this stage, 119 * such as MESA_SHADER_FRAGMENT. 120 */ 121 gl_shader_stage next_stage:8; 122 123 /* Number of textures used by this shader */ 124 uint8_t num_textures; 125 /* Number of uniform buffers used by this shader */ 126 uint8_t num_ubos; 127 /* Number of atomic buffers used by this shader */ 128 uint8_t num_abos; 129 /* Number of shader storage buffers (max .driver_location + 1) used by this 130 * shader. In the case of nir_lower_atomics_to_ssbo being used, this will 131 * be the number of actual SSBOs in gl_program->info, and the lowered SSBOs 132 * and atomic counters in nir_shader->info. 133 */ 134 uint8_t num_ssbos; 135 /* Number of images used by this shader */ 136 uint8_t num_images; 137 138 /* Which inputs are actually read */ 139 uint64_t inputs_read; 140 /* Which outputs are actually written */ 141 uint64_t outputs_written; 142 /* Which outputs are actually read */ 143 uint64_t outputs_read; 144 /* Which system values are actually read */ 145 BITSET_DECLARE(system_values_read, SYSTEM_VALUE_MAX); 146 147 /* Which patch inputs are actually read */ 148 uint32_t patch_inputs_read; 149 /* Which patch outputs are actually written */ 150 uint32_t patch_outputs_written; 151 /* Which patch outputs are read */ 152 uint32_t patch_outputs_read; 153 154 /* Which inputs are read indirectly (subset of inputs_read) */ 155 uint64_t inputs_read_indirectly; 156 /* Which outputs are read or written indirectly */ 157 uint64_t outputs_accessed_indirectly; 158 /* Which patch inputs are read indirectly (subset of patch_inputs_read) */ 159 uint64_t patch_inputs_read_indirectly; 160 /* Which patch outputs are read or written indirectly */ 161 uint64_t patch_outputs_accessed_indirectly; 162 163 /** Bitfield of which textures are used */ 164 uint32_t textures_used; 165 166 /** Bitfield of which textures are used by texelFetch() */ 167 uint32_t textures_used_by_txf; 168 169 /** Bitfield of which images are used */ 170 uint32_t images_used; 171 /** Bitfield of which images are buffers. */ 172 uint32_t image_buffers; 173 /** Bitfield of which images are MSAA. */ 174 uint32_t msaa_images; 175 176 /* SPV_KHR_float_controls: execution mode for floating point ops */ 177 uint16_t float_controls_execution_mode; 178 179 uint16_t inlinable_uniform_dw_offsets[MAX_INLINABLE_UNIFORMS]; 180 uint8_t num_inlinable_uniforms:4; 181 182 /* The size of the gl_ClipDistance[] array, if declared. */ 183 uint8_t clip_distance_array_size:4; 184 185 /* The size of the gl_CullDistance[] array, if declared. */ 186 uint8_t cull_distance_array_size:4; 187 188 /* Whether or not this shader ever uses textureGather() */ 189 bool uses_texture_gather:1; 190 191 /** 192 * True if this shader uses the fddx/fddy opcodes. 193 * 194 * Note that this does not include the "fine" and "coarse" variants. 195 */ 196 bool uses_fddx_fddy:1; 197 198 /* Bitmask of bit-sizes used with ALU instructions. */ 199 uint8_t bit_sizes_float; 200 uint8_t bit_sizes_int; 201 202 /* Whether the first UBO is the default uniform buffer, i.e. uniforms. */ 203 bool first_ubo_is_default_ubo:1; 204 205 /* Whether or not separate shader objects were used */ 206 bool separate_shader:1; 207 208 /** Was this shader linked with any transform feedback varyings? */ 209 bool has_transform_feedback_varyings:1; 210 211 /* Whether flrp has been lowered. */ 212 bool flrp_lowered:1; 213 214 /* Whether nir_lower_io has been called to lower derefs. 215 * nir_variables for inputs and outputs might not be present in the IR. 216 */ 217 bool io_lowered:1; 218 219 /* Whether the shader writes memory, including transform feedback. */ 220 bool writes_memory:1; 221 222 /* Whether gl_Layer is viewport-relative */ 223 bool layer_viewport_relative:1; 224 225 union { 226 struct { 227 /* Which inputs are doubles */ 228 uint64_t double_inputs; 229 230 /* For AMD-specific driver-internal shaders. It replaces vertex 231 * buffer loads with code generating VS inputs from scalar registers. 232 * 233 * Valid values: SI_VS_BLIT_SGPRS_POS_* 234 */ 235 uint8_t blit_sgprs_amd:4; 236 237 /* True if the shader writes position in window space coordinates pre-transform */ 238 bool window_space_position:1; 239 } vs; 240 241 struct { 242 /** The output primitive type (GL enum value) */ 243 uint16_t output_primitive; 244 245 /** The input primitive type (GL enum value) */ 246 uint16_t input_primitive; 247 248 /** The maximum number of vertices the geometry shader might write. */ 249 uint16_t vertices_out; 250 251 /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */ 252 uint8_t invocations; 253 254 /** The number of vertices recieves per input primitive (max. 6) */ 255 uint8_t vertices_in:3; 256 257 /** Whether or not this shader uses EndPrimitive */ 258 bool uses_end_primitive:1; 259 260 /** The streams used in this shaders (max. 4) */ 261 uint8_t active_stream_mask:4; 262 } gs; 263 264 struct { 265 bool uses_discard:1; 266 bool uses_demote:1; 267 bool uses_fbfetch_output:1; 268 bool color_is_dual_source:1; 269 270 /** 271 * True if this fragment shader requires helper invocations. This 272 * can be caused by the use of ALU derivative ops, texture 273 * instructions which do implicit derivatives, and the use of quad 274 * subgroup operations. 275 */ 276 bool needs_helper_invocations:1; 277 278 /** 279 * Whether any inputs are declared with the "sample" qualifier. 280 */ 281 bool uses_sample_qualifier:1; 282 283 /** 284 * Whether early fragment tests are enabled as defined by 285 * ARB_shader_image_load_store. 286 */ 287 bool early_fragment_tests:1; 288 289 /** 290 * Defined by INTEL_conservative_rasterization. 291 */ 292 bool inner_coverage:1; 293 294 bool post_depth_coverage:1; 295 296 /** 297 * \name ARB_fragment_coord_conventions 298 * @{ 299 */ 300 bool pixel_center_integer:1; 301 bool origin_upper_left:1; 302 /*@}*/ 303 304 bool pixel_interlock_ordered:1; 305 bool pixel_interlock_unordered:1; 306 bool sample_interlock_ordered:1; 307 bool sample_interlock_unordered:1; 308 309 /** 310 * Flags whether NIR's base types on the FS color outputs should be 311 * ignored. 312 * 313 * GLSL requires that fragment shader output base types match the 314 * render target's base types for the behavior to be defined. From 315 * the GL 4.6 spec: 316 * 317 * "If the values written by the fragment shader do not match the 318 * format(s) of the corresponding color buffer(s), the result is 319 * undefined." 320 * 321 * However, for NIR shaders translated from TGSI, we don't have the 322 * output types any more, so the driver will need to do whatever 323 * fixups are necessary to handle effectively untyped data being 324 * output from the FS. 325 */ 326 bool untyped_color_outputs:1; 327 328 /** gl_FragDepth layout for ARB_conservative_depth. */ 329 enum gl_frag_depth_layout depth_layout:3; 330 331 /** 332 * Interpolation qualifiers for drivers that lowers color inputs 333 * to system values. 334 */ 335 unsigned color0_interp:3; /* glsl_interp_mode */ 336 bool color0_sample:1; 337 bool color0_centroid:1; 338 unsigned color1_interp:3; /* glsl_interp_mode */ 339 bool color1_sample:1; 340 bool color1_centroid:1; 341 } fs; 342 343 struct { 344 uint16_t local_size[3]; 345 346 bool local_size_variable:1; 347 uint8_t user_data_components_amd:3; 348 349 /* 350 * Arrangement of invocations used to calculate derivatives in a compute 351 * shader. From NV_compute_shader_derivatives. 352 */ 353 enum gl_derivative_group derivative_group:2; 354 355 /** 356 * Size of shared variables accessed by the compute shader. 357 */ 358 unsigned shared_size; 359 360 /** 361 * pointer size is: 362 * AddressingModelLogical: 0 (default) 363 * AddressingModelPhysical32: 32 364 * AddressingModelPhysical64: 64 365 */ 366 unsigned ptr_size; 367 } cs; 368 369 /* Applies to both TCS and TES. */ 370 struct { 371 uint16_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */ 372 373 /** The number of vertices in the TCS output patch. */ 374 uint8_t tcs_vertices_out; 375 enum gl_tess_spacing spacing:2; 376 377 /** Is the vertex order counterclockwise? */ 378 bool ccw:1; 379 bool point_mode:1; 380 381 /* Bit mask of TCS per-vertex inputs (VS outputs) that are used 382 * with a vertex index that is NOT the invocation id 383 */ 384 uint64_t tcs_cross_invocation_inputs_read; 385 386 /* Bit mask of TCS per-vertex outputs that are used 387 * with a vertex index that is NOT the invocation id 388 */ 389 uint64_t tcs_cross_invocation_outputs_read; 390 } tess; 391 }; 392 } shader_info; 393 394 #ifdef __cplusplus 395 } 396 #endif 397 398 #endif /* SHADER_INFO_H */ 399